Collect file stats

[NikhilCollooru]

Collect the Partition - file stats (file name, size) from the HiveWriter to be stored in a TBD location and then later use it during scheduling to avoid the directory Listing call.
In this PR we are collecting the stats, tracking the blob size and then throwing it away.

depended by https://github.com/facebookexternal/presto-facebook/pull/1084

[highker]

comments on serde format

[highker]

The return format should be a page (or maybe a list of pages). Or if you wanna wrap it, it is also ok. But Page is the essential part for it. That allows the coordinator to write pages into manifests in ORC format.

Here is an example:

file_prefix	file_name	file_suffix	rows	bytes	col1_min	col1_max	col1_non_null_rows	col2_min	col2_max	col2_non_null_rows
ws://ws.atn5/abc/	1	.orc	1000	100030	ab	zzzd	900	0	15	1000
ws://ws.atn5/abc/	2	.orc	1200	110040	ac	zzf	1100	5	22	1200
ws://ws.atn5/abc/	3	.orc	1100	120050	bf	xwf	1000	19	42	1100

Each column types should be known beforehand. Or if not, types are serializable anyway, so they can be saved as List<TypeSignature>. For example, the above stats should come with column types: {varchar, varchar, varchar, bigint, bigint, varchar, varchar, bigint, bigint, integer, integer}.

Types and Blocks are the core data structure for Presto. So let's try to reuse them as much as possible.

[highker]

minor comments

[highker]

minor comments only